Staff Software Engineer

Doing product research online feels harder than ever. AI promised better information. Instead it flooded the web with hallucinations, generated content, and confident-sounding answers that have no idea what's true.
Product.ai is building the truth engine for shopping. When every other AI tool summarizes and guesses, we verify. Not opinions. Not averages. Ground truth.
Self-funded and profitable for over a decade, we're a team of e-commerce veterans who have been building commerce truth systems at scale. Over a billion dollars in annual transaction volume. Millions of customers served. Product.ai is our flagship - the new starting point for shopping in the AI era. We're in early alpha, building alongside our community of domain experts.
We need a Staff Software Engineer who owns the most critical technical decisions in the company: how we evaluate AI systems against reality, how we structure knowledge for both human and agent consumption, and how we keep the entire product rooted in verification instead of assumption. This is a builder-architect, not a staff engineer managing teams and process. Today, our senior architect is carrying platform-level decisions on top of building next-generation serving infrastructure. Your primary job is making the right architectural calls so the rest of the team can build with confidence.
Based in Santa Monica. Hybrid schedule with flexibility. For the right builder, we're open to remote.
What You Will Build

• Adversarial Verification Pipeline. Design and scale the systems that test AI models against reality, not benchmarks. You'll build the evaluation framework that cross-references marketing claims against specs, reviews, human network signals, and live transaction data. When an LLM says a product is "best in class," your pipeline determines whether that claim survives contact with evidence.
• Agent-Facing Protocol Layer. Ship the infrastructure that lets third-party AI systems consume verified commerce intelligence programmatically. REST APIs, JSON-LD structured data, MCP Server tools. You'll own the interface design between human-centered commerce and agent-centered commerce - where our defensive moat becomes an offensive revenue line.
• Knowledge Graph Infrastructure. Own the storage and query layer serving 75M+ product entities to three audiences simultaneously: humans (HTML), agents (JSON-LD), and internal systems (raw API). Same source of truth, polymorphic output, sub-200ms p99 latency. You'll decide how verified knowledge is structured, versioned, and served at scale - and how it degrades gracefully when upstream signals conflict.
• Evaluation Infrastructure. Build the machinery that runs continuous red-team verification against our own knowledge systems. Automated adversarial testing, human network signal integration, statistical rigor applied to confidence scoring. This is the internal immune system that catches our mistakes before users do.
• Foundational Architecture. Make the right calls on data models, services, queues, and caches across the entire stack. At a 20-person company, the staff engineer's architectural taste determines whether the team ships at 2x or 0.5x. Choose complexity only when complexity is earned.

Who You Are

• You have shipped AI to production and stayed to maintain it. You understand that evaluation and verification are harder than model training. Your opinions about RAG architectures, fine-tuning tradeoffs, and agentic system design are grounded in production incidents, not arxiv papers. You know the difference between a demo that impresses a boardroom and a system that serves real users without hallucinating.
• You think in systems, not features. You spot where a technical choice in the data model will compound into a philosophical problem in the product six months from now. You ask "what happens when this is wrong?" before you ask "how do we make this faster?" You build from first principles, not borrowed "best practices" from companies solving fundamentally different problems at fundamentally different scale.
• You are a builder who happens to be senior, not a senior who has stopped building. You write production code, not just architecture diagrams. At a 20-person company, the staff engineer reviews pull requests AND submits them. You are comfortable with ambiguity and you close it through iteration, not through longer planning cycles.
• You use AI to build AI. You have wired LLMs into your development workflow as force multipliers. You have a clear-eyed view of what they're good for (boilerplate, iteration speed, pattern matching) and where they fail (architectural decisions, novel tradeoffs, edge case reasoning). You are practical, not ideological.
• You value signal over process. You measure your work by what runs in production, not what lives in a design doc. You optimize for reality - does the code serve real users? - not internal metrics like coverage percentages or sprint velocity.

Ownership
We operate as a high-performance studio, not a typical corporation. Every builder at Product.ai is a partner, not an employee collecting a paycheck.

• Base: $350,000 - $450,000. Top of market. Your base covers your life; your ownership builds your freedom.
• Equity: You receive a formal ownership stake via Profits Interest Units (PIUs) - Class B Membership Interests with a $0 strike price. Not options. Not RSUs. Actual ownership that participates in the upside from day one and qualifies for Capital Gains tax treatment.
• Profit Sharing: You participate in the company's annual success. As an equity holder, you receive a pro-rata share of Free Cash Flow (FCF) - real cash, every year, not a promise tied to a future exit.
• Liquidity: We operate an Annual Tender Offer where the company buys back vested interests at fair market value. You can turn ownership into cash every year. No waiting for an IPO. No hoping for an acquisition.
• Benefits: 100% premium coverage for you and your family, daily catered lunches, and unlimited PTO that we actually expect you to use.

How to Apply
We skip the 60-minute recruiter screen. Instead, you'll answer a few short video questions so we can get to know how you think and work. Less time than a phone screen, and you can do it whenever works for you.
The Process:

Click Apply.

Record short video responses to our questions (no prep required - we want to see how you naturally think).

Your responses go directly to the hiring team. No ATS. No keyword parsing.

We're not looking for polished presentations. We're looking for signal on who you are.